The strong lottery ticket hypothesis (LTH) postulates that any neural network can be approximated by simply pruning a sufficiently larger network of random weights. Recent work establishes that the strong LTH is true if the random network to be pruned is a large poly-factor wider than the target one. This polynomial over-parameterization is at odds with […]
The Department of Bioengineering at the University of Pennsylvania and Dr. D. Kacy Cullen are pleased to announce the Doctoral Dissertation Defense of Justin Burrell. Title: Implantable Micro-Tissue Engineered Nerve Grafts to Maintain Regenerative Capacity and Facilitate Functional Recovery Following Nervous System Injury The public is welcome to attend virtually via Bluejeans.
Bone mineral density rapidly decreases upon withdrawal from intermittent parathyroid hormone (PTH) treatment despite its potent effect of promoting bone formation. To better understand this adverse phenomenon, this study first aimed to investigate the phenotype of PTH withdrawal in both intact and estrogen-deficient rat model by using a well-designed experiment combined with innovative longitudinal imaging […]
Abstract: "Heterogeneous catalysis is critically important in the chemical-processing and energy-related industries. Ideally, reaction sites on heterogeneous catalysts can be regenerated, and the catalysts should be active throughout the catalytic process. However, harsh reaction conditions and parallel side reactions can be harmful to the metal catalysts, making deactivation a common phenomenon. Finding a catalyst that […]
Abstract: Sintering is a severe problem with supported-metal catalysts in high-temperature applications, such as in automotive-emissions control, because it leads to a loss of catalytically active surface area. To stabilize the metal particles, in 2001, the Daihatsu group proposed using perovskite-supported metals, which they referred to as “intelligent” catalysts. The original goal was to regenerate […]
We present methods to sculpt light at the atomic and molecular scale to detect and control chemical transformations, en-route to improved planetary and personal health. First, we study plasmon-driven chemical transformations, focussing on the photocatalytic dehydrogenation of AuPd systems. Here, the Au acts as a plasmonic light absorber and Pd serves as the catalyst. Using […]
The Department of Bioengineering at the University of Pennsylvania and Dr. Ravi Radhakrishnan are pleased to announce the Doctoral Dissertation Defense of Lindsey Fernandez. The public is welcome to attend via Blujeans (Meeting ID 680 058 608 4). Title: "Multiscale Modeling of Cell Fate Switching to Predict Patient-specific Response to Combination Cancer Therapy"
Abstract: The lecture that I will cover is going to be on Hash Tables. The lecture is designed for an introductory CS course and/or a data structures course. We are in the last 3rd of the semester when this topic is covered. At this point, students are comfortable with programming (in Python). In addition, they […]
The Department of Bioengineering along with Drs. Walter Witschey and John Detre are pleased to announce the Doctoral Dissertation Defense of Eileen Hwuang. The public is welcome to attend via Zoom. Title: "Magnetic resonance imaging assessment of maternal uteroplacental hemodynamics during pregnancy" Zoom link Meeting ID: 745 723 0989 Passcode: 526101
Abstract: The fundamental issue underlying natural language understanding is that of semantics – there is a need to move toward understanding natural language at an appropriate level of abstraction in order to support natural language understanding and communication with computers.
Machine Learning has become ubiquitous in our attempt to induce semantic representations of natural language and support decisions that depend on it; however, while we have made significant progress over the last few years, it has focused on classification tasks for which we have large amounts of annotated data. Supporting high level decisions that depend on natural language understanding is still beyond our capabilities, partly since most of these tasks are very sparse and generating supervision signals for it does not scale.
I will discuss some of the challenges underlying reasoning – making natural language understanding decisions that depend on multiple, interdependent, models, and exemplify it using the domain of Reasoning about Time, as it is expressed in natural language.
